Tencent's AI team has released an exciting AI singing model called LeVo, which has sparked industry discussions with its powerful voice cloning, track generation, and high-fidelity music performance. It is reported that LeVo can rival leading industry models like Suno4.5 in multiple key metrics, securing a place for China’s AI music generation technology. AIbase has compiled the latest information to provide you with an in-depth analysis of LeVo's breakthrough features and potential.

On Par with Suno4.5: LeVo's Hardcore Strength

Developed by Tencent AI Lab, LeVo uses a language model (LM) architecture combined with LeLM and a music codec to generate hybrid tracks (merged vocals and accompaniment) or dual tracks (vocals and accompaniment separately). In terms of musicality, audio quality, harmony between vocals and accompaniment, and lyric alignment, LeVo surpasses existing open-source academic models. According to the latest evaluation, LeVo outperforms Suno4.5 by 0.21 points in lyric alignment capability (LYC), showcasing its excellent text control capabilities.

image.png

Project Link: https://levo-demo.github.io/

Zero-Sample Voice Cloning: A New Height in Personalized Music Creation

LeVo supports zero-shot voice cloning, accurately replicating target voice characteristics, including pitch, emotion, and rhythm, with just 3 seconds of audio. This function does not require large amounts of training data, significantly reducing the technical barriers for music creation. Whether customizing personalized voice tones for individuals or imitating famous singer styles, LeVo provides natural and smooth generation results, offering creators endless possibilities.

Track Generation: A Valuable Tool for Professional Music Production

Different from traditional AI music generation models, LeVo supports dual-track generation mode, enabling separate generation of vocal and accompaniment tracks, providing greater flexibility for post-production mixing and editing. This feature is particularly suitable for professional music producers, allowing them to easily achieve high-quality dual-track outputs and optimize the creative process. Compared to Suno4.5, which falls short in voice cloning and track support, LeVo sets a new benchmark in the industry.

High-Fidelity and Multi-Scene Applications

LeVo performs exceptionally well in audio quality, especially in musicality, harmony between vocals and accompaniment, and audio quality (MOS score). Although it slightly lags behind Suno4.5 and Mureka-O1 in song structure clarity, LeVo optimizes the generated results through multi-preference alignment methods, ensuring high-fidelity sound across various styles and scenarios. Whether it’s pop music, film scores, or advertising productions, LeVo offers professional-level outputs.

Open Source Commitment: Promoting the Development of the AI Music Ecosystem

Tencent has announced that LeVo will be released as open source, planning to provide complete code and pre-trained models for free use by developers worldwide. This move not only demonstrates Tencent's ambitions in the AI music field but also injects new vitality into the global music creation community. AIbase notes that LeVo’s open-source strategy will effectively lower the threshold for creation, helping content creators and music enthusiasts express their creativity.

The release of Tencent LeVo marks China’s AI music generation technology moving towards the forefront of the global stage. Its zero-shot voice cloning and track generation functions have brought revolutionary breakthroughs to music creation. Despite some gaps compared to Suno4.5 in certain indicators, LeVo, with its cost-effectiveness and open-source advantages, has become a strong competitor in the AI music field. AIbase believes that the launch of LeVo not only enhances China’s international influence in AI technology but also takes an important step toward democratizing music creation.